Conduit’s 172% Retention Puts Blue Yonder on Defense

The $6M Signal

Conduit, a San Francisco startup founded in 2023, closed a $6 million seed round, according to PRNewswire, in March 2026 led by Innovation Endeavors with participation from Y Combinator. The capital is modest by venture standards. The traction behind it isn't: more than 100 customers across North America, a net dollar retention rate of 172 percent, Conduit Software Inc reported, and zero logo churn. In enterprise software terms, those metrics mean that once a warehouse adopts the platform, it doesn't just stay; it expands aggressively across modules and facilities.

"We are building Conduit so any warehouse can run shipping and receiving on a single system. That means less time coordinating, higher throughput, and real visibility across the network," said Conrad Lilleness, founder and CEO. "It also means the structured workflows finally exist. You can't put agents on top of chaos. Before AI can direct warehouse operations, you need a system that knows what's actually happening."

That last sentence frames the real story. The logistics industry has been promised AI transformation for years. The bottleneck was never the models — it was the data. Most warehouses still run docks and yards on email, spreadsheets, radio chatter, and tribal knowledge. Critical operational data lives in inboxes and whiteboards, not in a structured system of record. Conduit's bet, and Innovation Endeavors' conviction, is that the company capturing that data at the source (the dock, the gate, the yard) becomes the substrate on which every subsequent AI workflow runs.

Harpinder Singh, managing partner at Innovation Endeavors, put it directly: "What excites us is that operators don't just adopt Conduit — they expand it across their entire networks. That combination of strong operational value and the team's execution makes us excited to partner with Conrad, Kevin, and the team."

The customer list reads like a cross-section of modern logistics: Flexport, Body Armor, Prism Logistics. These aren't pilots. Flexport integrated Conduit into its platform and replaced manual email chains with real-time visibility. Body Armor reported a 24 percent reduction in dwell time, the company's announcement found, and a 21 percent reduction in dock time within the first week. Prism Logistics, a 30-year-old 3PL with seven facilities, scaled throughput without adding headcount.

The product architecture reflects the expansion pattern. Conduit ships as modular components (dock scheduling, driver self-check-in, documentation, yard management) that warehouses can adopt one at a time. Each module feeds the same data layer. A site that starts with scheduling often adds yard management within months. That land-and-expand motion drives the 172 percent NRR.

Deployment speed reinforces the loop. Conduit claims customers go live in under two weeks, a timeline that legacy yard management systems (typically months-long implementations from vendors like Manhattan Associates and Blue Yonder) rarely match. The speed matters because it lowers the barrier to the first module, which starts the data flywheel.

The funding will deepen the product and grow the team. On the roadmap: predictive scheduling, real-time exception alerting, cross-facility benchmarking, and intelligent driver workflows — all built on the proprietary operational data Conduit now captures at scale. Engineering and customer success hiring will accelerate to meet demand from shippers and 3PLs racing to digitize.

The dock was the last place in the warehouse without a system of record. Now it has one. The question is what happens when the data layer is complete.

How the AI Dock OS Works

Conduit calls itself the operating system for shipping and receiving. That label describes architecture, not marketing. The platform sits between a warehouse's physical dock doors and the digital systems that run the rest of the supply chain (WMS, TMS, ERP, carrier portals) and replaces the phone calls, spreadsheets, and radio chatter that still coordinate most loading docks.

The product is modular by design. Three core modules (Dock Scheduling, Yard Management, and Driver Check-in) can be deployed independently or as a suite. A 3PL running only appointment scheduling can start there and add yard visibility later without re-implementing. Each module shares a single data layer, so a driver's check-in at the gate automatically updates the dock schedule, triggers a yard jockey move, and pushes timestamps back to the WMS. That unification is the product's central claim: one platform, one source of truth, no custom middleware.

Under the scheduling engine, the optimization problem is a multi-dock, parallel-machine sequencing challenge with stochastic arrival times, variable unload durations, and hard constraints — temperature-controlled doors, union labor rules, carrier appointments. Academic work published in November 2024 frames this exactly: minimizing makespan in cross-docking centers using an evolutionary algorithm paired with machine-learning-based hyperparameter tuning. That research (code available at github.com/yotitinogs/EA-ML) reports a 37 percent improvement in the GAP metric over state-of-the-art heuristics, a 30 percent gain over the PCH constructive heuristic for initial solutions, and a 21 percent lift from ML-based tuning versus static parameter sets, all within competitive compute times. Conduit has not publicly confirmed that this specific EA-ML stack runs in production, but the problem definition matches its domain, and the performance deltas explain why AI-native scheduling can outperform the rule-based engines incumbents still ship.

Yard management adds a spatial layer. The platform maps trailer positions in real time, supports drop-and-hook planning, and dispatches yard jockeys via mobile app. Driver check-in digitizes the gate: QR codes, photo capture for seal numbers and bill-of-lading exceptions, automated badge printing. All three modules feed a live dashboard that shows dock utilization, trailer aging, and labor allocation — data that previously lived on whiteboards or in supervisors' heads.

The stack runs cloud-native, API-first, with a mobile layer for gate and yard staff. Conduit's team, under 20 people as of its Y Combinator profile, ships updates on a weekly cadence. The net dollar retention suggests customers expand module adoption after initial deploy, a signal that the architecture supports growth without re-platforming.

What Customers Prove

Prism Logistics, a Northern California 3PL with 30-plus years moving food, beverage, and consumer packaged goods across seven facilities, provides the clearest window into what drives those retention numbers. Before Conduit, Prism's customer service representatives spent hours each day exchanging emails with carriers to confirm appointments, correct errors, and chase missing PO numbers. Appointments were logged manually at each site, creating conflicts and gaps. Drivers lined up at shipping windows; missing paperwork sent them back to dispatch and the end of the line. Centralized CSR teams had no real-time view of appointment status, delays, or volume trends across facilities, making it nearly impossible to dispute detention claims or staff proactively.

"We're always looking for tools that help us scale better, not just faster," said Jeremy Van Puffelen, Prism's president. "Conduit gave us a way to scale — without adding people or burning out the great team we have."

Prism evaluated several dock and yard management platforms and selected Conduit for its modular approach: scheduling, driver check-in, yard oversight, documentation, and analytics in one system across all sites. The results, documented in a G2 case study and confirmed by Prism's operations director Art Oliva on LinkedIn in October 2025, read like a before-and-after ledger:

Contactless QR check-in keeps drivers in their cabs. Digital records (time-stamped check-ins, chat logs, uploaded documents) give CSRs proof to dispute detention claims. A live network view across facilities lets supervisors monitor arrival patterns and volume trends for smarter staffing. Oliva noted the platform became a sales differentiator: "It is a sales point to our customers. Our customers are excited and love to hear that we are using a system like this."

Prism is not an outlier. Conduit's case study library names First Logistics, which consolidated scheduling, check-in, and documentation into one connected system; J.P. Logistics, which armed customers with self-serve, timestamped, photo-backed proof to defend against chargebacks; PS&L, which achieved efficient dock scheduling and improved customer service; and Encore, which doubled warehouse capacity through dock scheduling automation. Each story follows the same arc: manual, email-driven processes replaced by a single platform that scales without headcount.

The retention data compounds the case studies. A 172% NRR with zero churn suggests customers not only stay — they expand. In a market where incumbents like Blue Yonder and Manhattan Associates dominate enterprise yard management, that expansion signal is what attracted Innovation Endeavors and Y Combinator to lead the seed round.

Incumbents Fight Back

Blue Yonder has held the top quadrant in Gartner's Warehouse Management Systems Magic Quadrant for eighteen consecutive years — a streak that now extends into the 2026 edition released in June. That tenure gives the company a distribution moat Conduit cannot match overnight. But the same Business Wire announcement that touted the ranking also cataloged a flurry of AI additions: an AI-powered Warehouse Ops Agent that does real-time monitoring, dynamic forecasting, resource orchestration, and advanced slotting. Blue Yonder's yard management module now leans on camera-vision and machine learning to automate gate checks, generate 3D yard maps from mobile-mounted cameras, and eliminate RFID tags entirely. The company's own product page claims 100% automatic trailer identification and 100% reduction in gate and yard labor — figures that read more like aspirational targets than field-proven averages, but they signal where the incumbent is planting its flag.

Manhattan Associates, another Gartner Leader, positions Active Yard Management as a scheduling engine that optimizes inbound and outbound flows while maximizing yard utilization.

project44 takes a different vector. Its Movement platform, launched as an "AI-powered Decision Intelligence Platform for the modern supply chain," connects over 259,000 carriers for predictive ETAs and AI exception management across every mode. The company's yard module plugs into WMS, TMS, and ERP systems in real time, and integration partner Rayven advertises a two-to-twelve-week go-live window — faster than the traditional suite, slower than Conduit's claimed days. project44's heritage is visibility, not yard operations; it is layering yard management onto a tracking backbone rather than building from the dock outward.

SAP, meanwhile, embeds yard capabilities inside its broader logistics cloud.

The pattern across incumbents is clear: acquire AI capability (Blue Yonder bought One Network and Optoro), bolt yard modules onto existing suites, and market the breadth of the platform. Conduit's counter is modularity: each of its dock, gate, and yard modules can be adopted independently, and it has an integration library that spans WMS, TMS, and ERP layers via REST API, SAML/OAuth SSO, and bulk CSV. That library is the practical expression of an AI-native architecture: the platform was built to ingest heterogeneous data streams from day one, not retrofitted to do so.

For buyers, the choice crystallizes around two questions. First, do they want a yard module inside a WMS they already own, or a best-of-breed dock OS that talks to whatever WMS they run? Second, how much implementation risk they will absorb. Blue Yonder and Manhattan bring reference architectures at enterprise scale; Conduit brings 172% net dollar retention across 100-plus customers who expanded without adding headcount. The market is large enough for both models to coexist. But the talent implications diverge. Suite vendors hire integration architects and configuration specialists. AI-native platforms hire ML engineers who can train vision models on trailer footage, backend engineers who can normalize APIs, and product managers who can ship a new module in a sprint. The latter skill set is scarcer, and Conduit's fundraising is explicitly aimed at securing it.

The Talent War

The company's "AI-first" mandate rewrites the role definition. Their careers page states it plainly: "You'll be AI-first in every sense. You'll prototype faster than most engineers can write a ticket. You'll use AI to analyze operational data, accelerate customer discovery, and move at a level of speed and breadth that would have been impossible two years ago." That translates into three concrete hiring vectors. First, full-stack engineers who can build the AI layer atop Conduit's system of record: agents and features on top of structured operational data that doesn't exist anywhere else. Second, ML engineers who turn dock dwell time, gate throughput, and trailer-turn metrics into predictive models that schedule doors before a truck arrives. Third, automation engineers who bridge the software stack to physical yard equipment: RFID gates, camera-based check-in, yard-jockey dispatch tablets. The job specs call out AI tooling fluency as a daily requirement: "Access to any AI tools you need. We'll invest in whatever makes you faster" — and expect candidates to choose new technologies, not just adopt them.

Market data confirms the squeeze. Indeed listed 352 logistics AI openings as of the latest crawl. Mycelium Robotics characterizes the 2026 robotics hiring market as "accelerating demand and persistent supply constraints" across autonomous vehicles, humanoid robotics, warehouse automation, and defense, all fishing in the same shallow pool. Conduit competes for that talent against companies paying Anthropic-level bands ($205k–$562k median $395k) and Databricks-level bands ($140k–$319k median $250k). A warehouse automation engineer today needs fluency in AI, ML, and IoT sensor stacks — not as buzzwords but as production dependencies. The role blends classical controls (PLC ladder logic, safety-rated zones) with modern inference pipelines (TensorRT on edge gateways, ONNX model swap without downtime).

Operator roles are evolving in parallel. Conduit's customers (regional 3PLs like Prism Logistics running seven facilities, national shippers) need dock supervisors who read real-time dashboards instead of clipboard printouts. The platform's unified data layer closes the gap between WMS/TMS and the physical dock, delivering visibility that legacy systems never provided. That creates demand for operations analysts who can query the structured event stream Conduit produces: gate entry, door assignment, load/unload start, seal verification, departure. These analysts feed the AI training loop. Their annotations ("this dwell was weather, this was carrier no-show, this was labor shortage") become the labeled data that improves the next scheduling model.

The hiring signal extends beyond Conduit's headcount. The $6M seed round backed by Innovation Endeavors and Y Combinator, paired with 172% net revenue retention and near-zero churn, tells every other logistics AI startup that the talent bar just rose. Companies building on clean, proprietary operational data (not scraped web text) will attract engineers who want their models to touch steel. Conduit's stack proves that data exists. The next hire builds the agent that acts on it.

From Docks to Battlefields

The same problem Conduit solves at a loading dock, coordinating trucks, trailers, and labor in real time with fragmented data, appears at forward operating bases, on autonomous supply convoys, and in orbital transfer vehicles. The variables change: GPS-denied terrain, artillery-damaged roads, four-dimensional launch windows. The core challenge does not. Move assets from point A to point B when the plan has already fallen apart.

The U.S. Army's Project Sustainment makes this explicit. In July 2026 the service awarded prototype contracts to five firms (AM General, American Rheinmetall, Carnegie Robotics, HDT Robotics (operating as BLADE), and Stratom) to build medium autonomous platforms for tactical resupply. The goal: keep soldiers off high-risk routes while sustaining 24/7 logistics in contested environments. Field testing with operational units begins early 2027. The program absorbs the earlier M-MET effort, which itself grew from the squad-level S-MET robotic mule. As one Army official put it, "We have PLS systems (the larger tactical resupply trucks) and you have the SMET and there's really a gap between the two."

That gap is where commercial logistics AI enters. The TALUS contract ($15.9 million for a Tactical Autonomous Logistics Utility System) signals a shift. Military programs no longer build autonomy stacks from scratch. They license commercial driving software, adapt proven vehicle platforms, and layer on defense-specific requirements: operation without GPS, resilience to jamming, navigation after an artillery barrage has erased the road network. Einride and DAF Trucks announced a parallel track days before the TALUS award: SAE Level 4 autonomous electric freight targeting 2027 deployment. The commercial and defense timelines are converging.

Sensor companies are following the same path. Arbe Robotics, whose imaging radar detects small drones beyond 500 meters, won an Army contract for autonomous supply-chain trucks operating off-road alongside Forterra. "Physical AI needs reliable real-world perception," CEO Kobi Marenko said. "We are the sensor for that that can work in any weather, any lighting conditions." The same perception stack that guides a yard jockey through a crowded distribution center at 3 a.m. guides a robotic mule through a treeline under electronic attack.

Japan's smart robot market, valued at $40.1 million in 2025 and projected to $97.8 million by 2030, illustrates the global scale. Aging workforces drive automation in manufacturing, logistics, and elderly care simultaneously. The same demographic pressure hits U.S. trucking and military recruiting alike.

The 2030s forecast: widespread adoption of autonomous robotic logistics for base support, expeditionary resupply, and MEDEVAC. Swarm coordination across ground, air, and subsurface domains. Generative AI dynamically optimizing supply chains. The 2040s: quantum-accelerated path planning, robots repairing equipment and constructing infrastructure, blurred lines between domains.

Conduit's 172 percent net dollar retention proves the commercial engine works. The defense and space programs buying into that engine are not betting on vaporware. They are buying the same operational logic (visibility, automation, integration) hardened for environments where the dock door doesn't exist and the schedule is the enemy.

What Comes Next

The AI in logistics market hit $38.68 billion in 2026 and is projected to reach $180.63 billion by 2030 — a 47% compound annual growth rate. Generative AI in logistics, a sliver at $1.06 billion in 2026, climbs to $3.25 billion by 2030 at 32.3% CAGR. These figures from The Business Research Company and ResearchAndMarkets frame the runway Conduit is building on.

Conduit's product team frames its roadmap around the previously detailed capabilities, all built on that foundation. The convergence is clear: yard and dock operations generate the high-frequency, structured-unstructured data mix that agentic AI thrives on. Conduit's retention signals customers expand usage once hooked. The roadmap doubles down on that expansion, making the platform programmable by more personas, not just engineers. As the logistics AI market nearly quintuples in four years, the companies that own the execution layer at the dock will capture disproportionate value. Conduit's seed round bought them the lead. Their roadmap decides whether they keep it.

Working in AI? Zero G Talent tracks the openings: see every open Databricks role, browse AI jobs, openings at Anthropic, and the people building the field.